Hi Adrian, These are important questions – IMS services have to be very reliable and need to survive the loss of any server in the deployment. I’m going to call this ‘redundancy’.
Across the industry, different vendors have different ways of providing redundancy. Many of them provide 1+1 (or N+N for larger clusters) active-active or active-passive<http://www.sebastien-han.fr/blog/2012/05/26/differences-between-active-active-and-active-passive-cluster/> configurations, but the detail of how these work is different for different vendors; as far as I am aware, there is no industry standard. This applies equally to application servers and servers in the IMS core. Because we’ve designed Project Clearwater from the bottom up to handle this sort of problem, we think that it has a really good approach to redundancy, better than traditional 1+1 models. It requires fewer servers, so it is cheaper to run. It also allows us to cope better when something does go wrong, potentially dropping fewer calls. - Project Clearwater uses small, mostly stateless nodes, so when one fails, it has little impact. - It uses intelligent data stores, like Cassandra, to spread data so that any one node can fail with no loss of data. This means that Project Clearwater can run in an N+1 cluster, for example 3 Sprout nodes + 1 redundant node. This is a large saving over the 6 nodes needed for a 1+1 model. You can also run with N=1 if you want to run a 1+1 pair. For more detail on how this works, see the ‘Reliability and Redundancy’ section of http://www.projectclearwater.org/technical/clearwater-architecture/. If you are worried about redundancy, you should also think about what happens when a host fails in your virtualization infrastructure. It’s important to make sure that your virtual machines are spread out so that when a host fails, it only takes down one of your cluster. This is true for 1+1 and for Project Clearwater’s N+1 deployment models. VMware and OpenStack both provide anti-affinity rules as a way of making sure that machines are kept on different hosts. Finally, if you are looking for a top reliability IMS core, Metaswitch Networks produce a hardened, supported version of Project Clearwater called Clearwater Core. I would recommend taking a look at that if you are planning to go live with Clearwater. I hope that helps. Yours, Chris From: Clearwater [mailto:[email protected]] On Behalf Of Adrian Schwartz Sent: 24 April 2016 09:49 To: [email protected] Subject: [Project Clearwater] synchronization redundant IMS Hello This is a general understanding question. What is the standard way to implement 1+1 backup? How does the IMS servers/ AS servers are being backuped? What is the clearwater implementation of such a requirement? For example I have a master site which is having a list of users and list of provisioned information, what is the standard way to have this information forward to the backup server? Thanks Adrian
_______________________________________________ Clearwater mailing list [email protected] http://lists.projectclearwater.org/mailman/listinfo/clearwater_lists.projectclearwater.org
